Beaver on the Wisconsin Landscape: A Remote Sensing-Based Monitoring Framework and Multi-Scale Habitat Model
File(s)
Date
2020-12Author
Taylor, Janelle Megan
Publisher
College of Natural Resources, University of Wisconsin-Stevens Point
Advisor(s)
VanDeHey, Justin
Metadata
Show full item recordAbstract
Beavers (Castor canadensis) are the quintessential ecosystem engineer. Beaver dams and resulting impoundments can persist on the landscape for more than 150 years, even after site abandonment. A common method for quantifying beavers on the landscape uses aerial surveys to identify location and activity status of colonies based on food cache presence. Aerial surveys are both expensive and dangerous prompting management agencies to seek alternative monitoring approaches. Remotely sensed imagery such as aerial imagery collected by The National Agriculture Imagery Program (NAIP) is freely available and potentially useful for beaver monitoring efforts. Using NAIP imagery, I developed a supervised classification using 2011 and 2014 beaver aerial survey data provided by the Wisconsin Department of Natural Resources (WDNR) and defined a suite of aerial visual cues to aid in identification of beaver features on the landscape. Using these tools, I surveyed 84 watersheds and identified 921 areas of beaver features. The mean density estimate of features (x̄ = 0.119 features/km2) was significantly lower than the unadjusted estimate (x̄ = 0.154) reported by the WDNR, potentially indicating sampling bias. Additionally, I evaluated the efficacy of my remote sensing tools using accuracy assessments and observed a 23.6% increase to producer’s accuracy (P < 0.001). Implementation of remote sensing-based approaches to natural resources monitoring and management may provide a cost-effective alternative to current techniques, while simultaneously providing temporal “snapshots” of habitat use, beaver landscape alteration, and beaver feature persistence. Beaver (Castor canadensis) habitat use has been extensively studied, but rarely at multiple ecological scales. In fact, fewer than 25% of habitat-models evaluate habitat use at multiple scales. Beavers are unique in that they have a disproportionate effect on the landscape relative to allospecific species. They are capable of drastically altering both their immediate vicinity and surrounding landscape through creation of multiple dams, lodges, and felling behavior. To adequately understand an ecosystem engineer such as beavers, it is imperative to evaluate habitat use at relevant ecological scales. Here I present a modeling framework that examines beaver habitat use at a localized-scale (100-m and 200-m buffers) and watershed-scale using both agency-collected, and image-interpolated data. My models suggested that beaver habitat use was not dependent on ecological scale, but rather that my ability to discern beaver response to specified vegetative composition and structure was dependent on sampling scale. My results corroborated prior studies’ findings; beaver occurrence was negatively correlated with anthropogenic, rangeland, buckthorn and honeysuckle, central hardwood, and coniferous landcover classes, and was positively correlated with stream density, proximity to known beaver locations, aspen, and wetland landcover classes. To my knowledge, this is the first study to examine beaver habitat use at multiple relevant ecological scales.
Subject
Beaver
Castor Canadensis
Habitat Model
Remote Sensing
Supervised Classification